InstituteLeibniz Institute for Economic Research

at the University of Munich

ICT and productivity: a roadmap for empirical research

The department for “Industrial Organisation and New Technologies” at the Ifo Institute,

headed by Prof. Tobias Kretschmer, focuses its research on strategic issues in ICT markets.

Along with studies of the contribution of ICT to productivity at company, sector and country

level, the department is especially concerned with the complex interactions between different

information technologies, and also with complementarities between ICT use and organization

structures. Other lines of research are the dynamics of consumer electronics markets and the

diffusion of new communications technologies.

Press contact: Annette Marquardt, tel.: 089/9224-1604, e-mail: marquardt@ifo.de.

The underlying studies can be accessed on the website of the ifo department of Industrial Organisation

and New Technologies: http://www.cesifo-group.de/int

© Ifo Institute – Leibniz Institute for Economic Research at the University of MunichImage: Deutsche Telekom AG, Bonn.

We would like to thank Deutsche Telekom for its financial support to this research project.

1

The European growth model enjoyed decades of

success and brought prosperity to several million

people in Europe. Now the European debt crisis has

made it clear that this growth model has reached

its limits: structural problems followed by a halt in

productivity growth are seen as the main causes

of the difficult economic conditions in some parts

of Europe. A comparison with the USA shows that

Europe as a whole has some catching up to do –

since the mid-1990s, productivity growth in the USA

has been ahead of that in Europe. The big advances

in productivity in the USA have been mainly

achieved by the extensive use of information and

communications technologies.

Management summary

– Although the positive contribution of information and com-munications technology (ICT) to productivity and eco-nomic growth is beyond dispute, the academic and pol-icy-related literature is marked by widely varying findingsand approaches to measuring the contribution of ICT.

– Most approaches are based on a production function, inwhich inputs (production factors) are transformed into out-puts (goods produced). ICT is included as an additionalproduction factor. Growth accounting compares the doc-umented investment in ICT with the goods produced.These studies are mostly at the sectoral level. Econo-metric estimates are designed to quantify the additionaloutput generated by increased investment in ICT. Thismethodology is especially applied at the firm level.

– Growth accounting studies distinguish between the ICTcontribution in a narrow sense (i.e., attributable to ICTinvestment alone) and the contribution of the knowledgeeconomy (which also includes the indirect effects of ICTon other sectors and production processes). Up to 80%of the growth in productivity over the last few decades isattributed to the knowledge economy.

– The vast majority of econometric estimates finds that ICThas a positive effect on growth. On average, the studiesconsulted show that a 10% increase in ICT investmentresults in a growth in output of 0.5–0.6%. In the last fewyears, this growth effect has actually been higher.

– This rising growth effect is consistent with the increasedability of organizations to make optimum use of infor-mation and communications technology. It is thereforeparticularly interesting to know which complementaryinvestments and capabilities contribute to the optimumuse of ICT.

In the academic and policy-related literature, widely differ-ent figures are put forward for the relationship between in-formation and communications technology (ICT) and pro-ductivity. The aim of this report is to produce a roadmapfor empirical research, giving a brief account of the key meth-ods of measuring productivity followed by an overview ofthe spread of existing estimates of the effect of ICT on pro-ductivity.

1. Productivity as a driver for prosperity

Productivity in its most general sense describes the relation-ship between the output produced and the inputs used todo so. Where there is information on the individual values,the relationship can be established and measured both atthe macroeconomic level for a national economy and at com-pany and industry level.

Productivity measurement serves not only to assess the ef-ficiency with which input factors are transformed into out-put but also to provide essential clues to prosperity andstandard of living within a national economy. Because, asproductivity increases, wages and salaries can rise too.Studies of productivity and its growth are therefore of in-terest not only to economists; they also provide importantpointers for the development and sustainability of socialprogress.

The drivers for productivity growth are continuous improve-ment in the quality of the input factors, such as the level oftraining and technological knowledge, and also the advanceof product and process innovation. The more productivelyinput factors can be deployed, the greater the return on in-vestment in physical and human capital and the higher thestandard of living in a country. Social progress is thereforeclosely linked to productivity growth and reflects not only theefficient provision of output values but also the accumula-tion of intangible assets such as human capital, knowledgeand access to knowledge networks.

International comparisons offer interesting insights intothe performance of national economies. If we comparethe trend in productivity growth between the economiesof the USA and Europe over the period from 1978 to 2007,some clear differences can be discerned (see Figure 1).Whereas productivity growth in Europe ran ahead of thetrend in the USA up to the early 1990s, the reverse hasbeen true since the mid-1990s. Instead of a progressiveconvergence between the two economic blocs, there wasa divergence in productivity growth. The USA has twice ex-perienced accelerating growth in productivity, in the peri-ods after 1995 and after 2000. Europe, on the other hand,suffered a significant slowdown in growth in the same pe-riod. It is only since 2006 that the figures show a converg-

Tobias Kretschmer, Mélisande Cardona, Thomas Strobel

ICT and productivity:a roadmap for empirical research

ICT and productivity2

ing trend between the two regions. If we take the initiallyfavorable productivity developments in Germany and Francein 2005/2006 as representative of the EU-15 countries,these saw a massive collapse in the course of the financialcrisis in 2008/2009. In contrast, development in the USAwas much more positive, and we can once more see a di-vergent trend in growth rates between the two regions (Con-ference Board, 2011).

To explain the weak growth in Europe, we can identify notonly the contributions from the number of hours worked andimproved quality in the input factors, but also other factorssuch as research and development spending and the in-creased use and diffusion of information and communica-tions technology (ICT). The ICT element of total capital ex-penditure itself differs markedly between the regions (seeFigure 2). Since the early 1980s, the proportion of ICT cap-ital invested in Europe and the USA has diverged sharply.This tendency intensified more and more in the mid-1990s

and only stabilized when the dot.com bub-ble burst in the period after 2000.

When deriving productivity measures, dif-ferent concepts need to be applied accord-ing to the question asked and the availabil-ity of the necessary data. As an output mea-sure, we generally use the production val-ue, or the production value adjusted for in-termediate inputs, i.e., what is known asgross value added. Along with various out-put measures, different input measures mayalso be considered, with a distinction be-tween univariate and multivariate produc-tivity measures, according to whether weare investigating the relationship betweenone or more inputs and the output pro-duced.

The transformation of input factors into outputs can be em-pirically measured and presented in various ways. If we mea-sure output as gross value added and apply the most com-mon version of a Cobb-Douglas production function, we ob-tain the following relationship:

Y = AKα Lβ

Equation 1: Cobb-Douglas production function

Equation 1 assumes a multiplicative relationship betweenthe output produced Y and the input X needed to generateit. K then represents the capital invested in the productionprocess, and L is the labor needed, while α and β stand forthe respective factor elasticities in the input factors. The fac-tor elasticities show the percentage by which output risesgiven a 1% increase in the input. Finally, the value A mea-sures the total factor productivity (TFP), which is often tak-en as a measure of technological progress.

The TFP assumes particular importance inconnection with the influence of ICT as ameasure for a general purpose technology(GPT; Bresnahan and Trajtenberg, 1995) oras a basis for the ‘spill-over’ theory of ICT.The first suggests that ICT constitutes aspecial kind of technology, which can bedeployed to enhance efficiency in manyproduction processes. It is not only the di-rect use of ICT but also the development ofnew complementary technologies thatleads to productivity increases. The spill-over theory assumes that initial productiv-ity increases in ICT-producing industriesreach other industries, particularly thosethat make intensive use of ICT, after a cer-tain time delay.

-1

0

1

2

3

4

1978 1982 1986 1990 1994 1998 2002 2006

Labor Productivity Growth

Source: EU KLEMS (2009).

Productivity Growth in the USA and Europe (EU-15), 1978–2007

EU-15

USA Period Averages

%

Figure 1

0

2

4

6

8

10

12

14

16

18

1980 1983 1986 1989 1992 1995 1998 2001 2004 2007

ICT Share of Total Capital Cost

Source: EU KLEMS (2009).

ICT Intensity in the USA and Europe (EU-15), 1980–2007

EU-15

USA

%

Figure 2

ICT and productivity 3

2. How the contribution of ICT to productivity ismeasured: growth accounting and econometricestimates

2.1. Growth accounting: causes of productivity growth

Growth accounting breaks down productivity growth into itsinput elements. Here, both output and input values are gen-erally related to the labor factor (Jorgenson et al., 2005, 2007;Aghion and Howitt, 2007) and formulated as a dynamicequation, i.e., the growth rate for labor productivity (Δln y) isexplained by the growth rates for the input factors. In thisapproach, a distinction is made between the capital inten-sity of ICT (Δln kICT) and non-ICT goods (Δln kNICT), and la-bor quality (Δln LQ) is taken as an explanatory factor for la-bor productivity. The growth contribution of ICT can be fur-ther isolated by looking separately at thegrowth in total factor productivity in ICT-in-tensive (Δln AICT) and non-ICT-intensive indus-tries (Δln ANICT) (see equation 2).

Equation 2: Input contributions in growth accounting

In order then to calculate the contributionsto productivity growth, the growth rates forthe individual factor inputs are weighted withtheir nominal factor income percentages.These describe the quantitative contributionof the two input factors, capital ( –υ ICT and

–υ NICT) and labor ( –υ L), to the output produced. In order toweight the two TFP values, we use the nominal gross val-ue added percentages for the industries ( –ω ICT und –ω NICT).The two figures below show a breakdown of the aggregat-ed growth in labor productivity into its individual contribu-tions for the USA, as derived from equation 2. Instead ofreported annual contributions, averages over three periodsare shown.

It can be seen from Figure 3 that the growth in labor pro-ductivity in the USA during the New Economy period(1995–2000) in particular was driven by increased ICT cap-ital intensity. Where the greater contribution to growth be-tween 1973 and 1995 came from non-ICT factors, after1995 there was an increased substitution of ICT for non-ICT goods. The 0.58% TFP growth of ICT-intensive sec-tors also made a major contribution to productivity growth.Although the ICT contribution in the USA has remained highsince 2000, it is clear that the productivity effects associ-ated with ICT are waning.

Similar findings emerge when aggregating individual growthcontributions to a shared contribution from ICT capital in-tensity and TFP for the ICT-intensive industries, and thesame contributions taking account of labor quality andthe TFP contributions of non-ICT producing industries (theknowledge economy), are aggregated (see Figure 4). Thefirst form of aggregation is equivalent to the ICT contribu-tion in a narrow sense, without any spill-over effects. The“knowledge economy”, on the other hand, measures awide range of effects associated with the use of ICT. Hence,it also includes the effect of an altered composition of theworkforce, with more highly-qualified workers (complemen-tarities between better qualified workers and ICT) and themacroeconomic TFP contribution, i.e., the contribution ofall industries. The cross-industry TFP contribution in par-ticular is a major feature of the technological progress and

Output and productivity quantities (Schreyer and Pilat, 2001)

In contrast to the pure production value (gross output), gross value added has the advantage of avoiding double-counting by cleansing the figures of inter-mediate inputs. Measurement on the principle of gross value added also offers a direct link to income per capita or per hour, which are widely used indicators to measure the standard of living. In the case of univariate productivity measurements, the output is related to one input factor, usually in the form of labor productivity (output to labor). It is also possible to relate it to the factor of capital (capital pro-ductivity). With multivariate productivity measures, we look at output in relation to all factor inputs. The re-sulting indicators are then called total factor pro-ductivity (TFP) or less commonly multi-factor produc-tivity (MFP).

QLNICTNICTICTICT Lkky lnlnlnln

NICTNICTICTICTQ AAL lnlnln

0.25 0.19 0.31

0.4

1.010.58

0.45

0.490.690.25

0.580.38

0.14

0.42

0.54

1.49

2.692.50

0.0

0.5

1.0

1.5

2.0

2.5

3.0

1973–1995 1995–2000 2000–2006

Non-ICT TFP

ICT TFP

Non-ICT CapitalIntensity

ICT CapitalIntensity

Labor Quality

Sources of Productivity Growth in the USA

Growth Contributions in %

Source: Jorgenson et al. (2008).

Figure 3

ICT and productivity4

innovative activity of a knowledge-based society, as wastriggered by ICT in the New Economy period. Here again,we see the impact of an increased focus on technologyon productivity growth in the USA, especially in the peri-od 1995–2000.

Growth accounting captures the influencing fac-tors documented in official statistics, which con-tribute to the growth in labor productivity. Growthaccounting allows the growth effects to be relatedto ICT intensity in a narrow sense and to the knowl-edge economy in the broader sense. The latter thenalso reflects the indirect growth effects of increaseICT use.

2.2. Econometric estimation of a production function

As with growth accounting, the Cobb-Douglas function isagain the basis for econometric estimation, for which it takesthe following form (Brynjolfsson and Hitt, 1995, 2003):

Equation 3: Logarithmic Cobb-Douglas production function

Here, the output Q (in company and industry studies, usu-ally taken as gross value added; in country studies, gener-ally GDP) is related to the ICT capital C, the capital stock ofnon-ICT goods K and labor L. The indices i and t representthe unit of analysis (i.e., company, industry or country) andthe time respectively. The existing ICT capital is sometimesapproximated by penetration rates; in some studies, for ex-ample, telecommunications capital is derived from the num-ber of telephone lines per inhabitant, and IT capital at firmlevel is taken as the number of computers per employee.Common control variables are time and dummy variablesfor the unit of analysis concerned (fixed-effect models), thelatter capturing constant idiosyncratic productivity effects –for example, some companies are constantly more pro-ductive, e.g., because of good management or an advan-tageous market position.

Unlike growth accounting, econometric estimation estimatesthe relative contributions to growth of the individual inputs asparameters rather than calculating them from income statis-tics. In contrast to growth accounting, econometric estimatescan be used to identify statistically significant and causal re-lationships. That means, for example, that conclusions canbe drawn as to the scale of the growth contribution basedon a number of observations of individual companies.

A particular challenge is how to interpret anycausal relationships. For example, invest-ment in ICT may be both a cause and aneffect of economic growth. This is referredto as the endogenity problem, as the direc-tion of causality cannot be established be-yond doubt. In order to circumvent the prob-lem of reverse causality, some empiricalstudies use different econometric methods,in which the ICT variables are replaced byspecified time-delayed variables (e.g., Bloomet al., 2010, Brynjolfsson and Hitt, 1995,Hempell, 2005b, Tambe, 2011). As currentgrowth cannot have any influence on pastinvestment, we can rule out reverse causal-ity in this case. Alternatively, structural mod-els can be estimated, in which the differentinfluences are modeled in multiple equations

Total factor productivity (TFP) The total factor productivity (TFP) derived as part of the production function is calculated as the residual value of output, capital and labor. It measures how productively the individual input factors are combined with each other to generate the output. Under some assumptions, TFP can then be interpreted as tech-nological progress. Where there are variances from the assumptions of constant economies of scale, ef-ficient production and competitive factor markets, TFP measures not only non-constant economies of scale and changes in production efficiency but also chan-ges in capital utilization and cyclical effects. If we look at different groups of goods and products, with each group weighted with appropriate quality-adjusted prices, any form of measurable technological pro-gress will already be contained in the (priced) input values. We then speak of embodied technological progress. On this basis, the residual TFP is then inter-preted as disembodied technological progress.

0.43

0.59

0.38

0.70

0.81

0.72

0.0

0.2

0.4

0.6

0.8

1.0

1973–1995 1995–2000 2000–2006

senseICT in anarrow

KnowledgeEconomy

Productivity Growth Contribution of ICT and Knowledge Economy in the USA

Growth Contributions in %

Source: Jorgenson et al. (2008).

Figure 4

itititit LKCQ lnlnlnln 321

itit controlsLln3

ICT and productivity 5

(Koutroumpis, 2009, Röller and Waverman, 2001). Anoth-er approach is to start by estimating broadband penetra-tion using a diffusion equation which is unaffected by cur-rent economic growth (Czernich et al., 2011).

A further econometric approach involves estimating TFPregressions (Basu et al., 2003, Inklaar, 2008, Stiroh, 2002a).In this case, TFP, which may sometimes be interpreted asan efficiency gain from the use of all input factors (see boxon TFP), forms the dependent variable explained by ICT. Thisapproach thus offers a test of whether ICT use producesspill-over effects that are not captured by (ICT) capital in-tensity. As spill-over effects mainly manifest themselves insecond-round effects, this approach looks mainly at the sta-tistical significance of the delayed impact of ICT. Of partic-ular interest here are the delay with which the spill-over ef-fects occur and the existence of complementary investments(e.g., management quality). Both factors are often difficult tocapture empirically.

Apart from the approaches mentioned above, there are oth-er ways of quantifying the economic value of technology.One thread looks at the effects of ICT investment on profitabil-ity and so attempts to evaluate the role of ICT in creating strate-gic competitive advantage (e.g., Im et al., 2001, Tam, 1998).Consumption theory is also the origin of the approach of cal-culating the consumer return, which quantifies in euros theamount by which consumers benefit from new technologies(Hausman et al., 1997, Greenstein and McDevitt, 2009).

Estimating production functions is a more flexiblealternative to growth accounting. It can be used todraw concrete, causal conclusions about the contri-bution of an increase in ICT capital to productivitygrowth. However, reliable results can only be expect-ed from studies that address the problem of the pos-sible effect of productivity growth on ICT investment.

3. Overview of the empirical evidence for thecontribution of ICT to productivity

3.1. Initial summary

The abundance of studies on ICT and productivity need tobe differentiated not only according to the methodological ap-proach taken but also according to other criteria. A major as-pect is the ICT product being examined. Various studies fo-cus exclusively on communications technology, and thesecan in turn be broken down into studies of data communica-tion (mainly broadband), and voice telephony (both fixed andmobile networks). Another category is made up of studies ofinformation technology (computer hardware and peripherals).It is only since the price deflators have been enhanced andsoftware included as investment in the national statistics thatsoftware has increasingly been considered also. Another im-portant differentiating feature is the level of aggregation ofthe units of analysis – companies, industries or countries.

Table 1 provides an overview of these differentiating featuresand names the two most-cited papers in their respective fields(measured by the Google citation index). As can be seen fromthe Google citation index, the focus of the existing researchis mainly on examining the deployment of information tech-nology in companies and, at the country level, of ICT in gen-eral. The field of communications technology is dominatedby econometric studies, while growth accounting studies arefavored for IT and ICT measurements. The latest field is con-cerned with the productivity effect of broadband, but thereare as yet no studies of the use of voice telephony at the firmlevel or data communication at the industry level.

The existing literature on the growth contribution ofICT encompasses different levels of aggregation andindividual technologies. Growth accounting is usedespecially at the macro level, and production func-tions at the micro level.

3.2. Growth accounting results

Figure 5 compares the growth rates for weighted ICT cap-ital intensity for different growth accounting studies. De-spite the use of standardized methods and largely identicaldata sources, the results differ in detail. Along with differentperiods covered, the reasons lie in the use of different pricedeflators and returns on capital, the latter being incorporat-ed into the weightings of the factor income elements. Somestudies also relax the restrictive assumptions by consider-ing capacity utilization and non-competitive factor markets(e.g., Oliner et al., 2007).

As the comparison between the USA (blue lines) and theEU (red lines) in Figure 5 shows, the EU invested much less

Factor elasticity and marginal product By estimating the Cobb-Douglas production function, we can use the estimated parameter β1 in equation 3 to obtain the elasticity of the factor input ICT capital directly. This indicates the percentage by which output changes given a one-percent increase in ICT capital. To measure the amount in euros by which output changes given a one percent increase in ICT capital, we need the marginal product of ICT capital (GPICT), which can be calculated by reformulating the elasticity using Q (output) and C (ICT capital) as follows:

The marginal product indicates the effect of the last euro of ICT capital spent.

ICT and productivity

in ICT in the late 1990s, which is however less strongly re-flected in the contribution of ICT capital intensity becauseof low labor productivity in the same period. In the 2000s,investment tails off in both regions.

The next two charts (see Figure 6) show the weighted TFPgrowth rates. As mentioned earlier, TFP is an important in-dicator as it reflects efficiency gains. Here we can clearlysee the much greater significance of the ICT-producing in-dustry in the USA, as all US-based studies find a highergrowth contribution from ICT than their European counter-parts. For aggregated TFP across all sectors, it can be seenthat since 1995, TFP has been much higher in the USA thanin Europe. However, the latest figures published by the European Commission again show a trend towards pre-1995 productivity levels.

Further differences between Europe and the USA emergewhen we break down the growth in labor productivity bysector (see Figure 7). This shows that the USA has pro-ductivity advantages not only in ICT-pro-ducing but also in ICT-intensive (i.e., heavyICT-using) sectors such as the financial in-dustry, research-intensive sectors and cor-porate services. The EU, on the otherhand, has higher growth rates in non-ICT-intensive sectors. Given this successfuluse of ICT in the USA, we may ask whetherany complementary investments, such ashard-to-quantify organizational capital andmanagement quality, have not been cap-tured. In one of the few studies on this,Bloom et al. (2010) highlight managementquality to explain differences in ICT pro-ductivity between the UK and the USA.Acharya and Basu (2010) use intangiblecapital, which reduces residual TFP.

If we look at the contribution of the knowl-edge economy (ICT in the broader sense)

6

to labor productivity from various sources for the USAand Europe (see Table 2), similar contributions can be foundfor both regions. However, the USA registers a much widerspread of positive growth contributions in the period1995–2005. In terms of ICT contribution, on the other hand,Europe lags significantly behind the growth contributionsin the USA right up to the end of 2000. Only after 2000do these start to converge, while the ICT contribution inEurope displays a downward trend.

Growth accounting studies in Europe and the USAshow a large element of ICT in productivity growthin the last few decades. The growth contribution ofICT in the USA was generally higher than in Eu-rope, but particularly the ICT growth contributionin the broader sense (i.e., via a macroeconomic tech-nology transfer within an advancing knowledgeeconomy) has caught up in Europe in the last fewyears.

Table 1 Star paper matrix

ICT product/ aggregation level

Company level Industry level Country level

Voice telephony (mobile/fixed network)

– Greenstein/Spiller (1995, 44); Correa (2006, 14)

Röller/Waverman (2001, 441); Hardy (1980, 184)

Data (Internet/broadband)

Grimes et al. (2009, 6) – Lehr et al. (2006, 73); Crandall (2007, 58)

IT (hardware/software) Bresnahan (2002, 1609); Brynjolfsson/Hitt (1996, 1319)

Baily/Lawrence (2001, 241); Stiroh (1998, 133)

Gordon (2000, 1123); Jorgenson (1999, 298)

ICT (general) Bertschke/Kaiser (2004, 118); Hempell (2005, 92)

Stiroh (2002b, 608); Morrison (1997, 251)

Oliner/Sichel (2000, 1417); Jorgenson (2001, 991)

Note: The number of Google Scholar citations is shown after the year of publication. Source: Google citation index (spring 2011).

Growth of ICT Capital Intensity and in Percentage of Labor Productivity in the USA and EU, 1970–2008

0.0

0.2

0.4

0.6

1970 1980 1990 2000 20100.0

0.5

1.0

1.5

1970 1980 1990 2000 2010

Year

Weighted Growth in %

USA EU-15

Year

Share in %

Figure 5

Sources: USA data: Cette (2009), Gordon (2010), Jorgenson (2001, 2007, 2008), Oliner (2007);EU data: EUKLEMS (2008), EC (2011), O’Mahoney (2009), Timmer (2003), Van Ark (2002, 2008),Vijselaar (2002).

ICT and productivity

3.3. Findings on the output elasticity of ICT

The second major thread in the literature on ICT and pro-ductivity assesses the impact of increased ICT capital onoutput (parameter β1 in equation 3). In the usual modifiedCobb-Douglas specification, this effect is measured as apercentage change. This means that β1 indicates howmuch output would increase if ICT investment were raisedby 1%. In the literature there are a number of variants fromthis specification, which are not considered below for easeof comparison. As most studies mention several elastici-ties, the comparison in Figure 5 always takes the estimateproduced by the most conservative method (e.g., fixed-effects estimate, endogenity-checking estimation meth-ods). A list of the studies referred to can be found in theAppendix.

The histogram in Figure 8 shows a cluster-ing of the estimated elasticities around the0.05 range with a few positive outliers. Thereare also a few negative elasticities. Also, theright-hand chart shows that the elasticitiesincrease over time with the accompanyingincrease in the ICT capital stock.

The estimates of elasticity in the ICT prod-uct/aggregation level matrix in Table 3 showno systematic differences in elasticities ac-cording to aggregation or product level. In-terestingly, differentiating by region – unlikein growth accounting studies – producesno significant differences. Overall, the stud-ies of the output effects from the deploymentof ICT capital yield a consistent picture, witha mean of 0.05–0.06. That then means thatan increase of 10% in ICT investment resultsin output growth of approx. 0.5–0.6%. Itshould be stressed that most of the studies

examined address the problem of reverse causality.

Much more ambivalent are the results of the TFP regres-sions and other empirical studies of the GPT hypothesis.Brynjolfsson and Hitt (2003), Baily and Lawrence (2001) andBasu (2003) find positive empirical evidence for the existenceof spill-over effects and for ICT as GPT for the USA. Sometake productivity comparisons between industries as indirectverification of the spill-over theory (Bosworth and Triplett,2003, Stiroh, 2002), while others see no evidence for this atall (Inklaar et al., 2008, Van Ark and Inklaar, 2005). Other stud-ies use various descriptive statistics to compare ICT with pre-vious ground-breaking technologies such as electricity. WhileCrafts (2002) and Jovanovic and Rousseau (2005) stress thatICT need not fear comparison with other major technologi-

cal innovations, Gordon (2000) draws rathera skeptical conclusion as to the GPT char-acter of ICT. Despite the abundance ofanecdotal evidence of innovations arisingfrom the use of ICT (Brynjolfsson and Saun-ders, 2010), and the assertion that “the sus-pect’s fingerprints are all over the crimescene” (Basu et al., 2003), there is as yetno definite empirical proof of the GPT hy-pothesis.

In the existing productivity studies,the mean output elasticity of ICT in-vestments lies between 0.05 and0.06: a 10% increase in ICT invest-ment then raises output by 0.5–0.6%.This figure has tended to increasein the last few years.

7

Weighted TFP Growth in the USA and EU, 1970–2008

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1970 1980 1990 2000 20100.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1970 1980 1990 2000 2010

Year

%

USA EU-15

Year

In the ICT Sector Across All Sectors %

Figure 6

Sources: USA data: Cette (2009), Gordon (2010), Jorgenson (2001, 2007, 2008), Oliner (2007);EU data: EUKLEMS (2008), EC (2011), O’Mahoney (2009), Timmer (2003), Van Ark (2002, 2008),Vijselaar (2002).

8.1

1.2 0.3

10.0

4.7

-0.2

5.9

2.0 2.1

7.5

1.91.0

-2

0

2

4

6

8

10

12

ICTproducing

ICTintensive

Non ICT ICTproducing

ICTintensive

Non ICT

US EU

Labor Productivity Growth by ICT Sectors

Labor Productivity Growth in %

Source: Inklaar et al. (2003).

1990–1995 1995–2001

Figure 7

ICT and productivity

4. Summary and questions for the future

The roadmap set out here makes it clear that the contribu-tion of ICT to productivity growth in individual companies,sectors or national economies is independent of the methodof measurement or the level of aggregation. Both growth ac-counting and productivity studies arrive at findings that as-sign a leading role to ICT in securing sustainable growth.

Still, there is no prevailing consensus in theliterature on a number of questions:

What is the significance of complementaryinvestments, particularly in the form of intan-gible (non-measurable) capital?

The existing studies show that some com-panies are able to deploy ICT more produc-tively and hence to achieve a higher growthcontribution from ICT. The important task is now to research the reasons for this fur-ther, and particularly to identify and quanti-fy the “soft” organizational and strategic factors. Well-founded findings on the majorcomplementary functions could be highlyrelevant to companies as well as policy-makers if we are to make optimum use ofICT investments and retain them as driversfor commercial success and productivitygrowth.

Is the ICT-producing industry driving pro -gress, or is it the diffusion of these technolo-gies into other sectors? Is it embodied or dis-embodied technological progress?

The ICT-producing industry is making con-stant progress in developing and marketingnew information and communications tech-nology, and so constitutes a key driver forprogress and growth. At least as interest-ing, however, is the question whether growth

can be further stimulated by the diffusion and use of ICT (the“knowledge economy”). Confirmation of this theory, partic-ularly at the sectoral level, would then equate the growthcontribution of ICT with that of a “general purpose technol-ogy”, serving as a stepping stone to further innovation.

What contribution can we expect from ICT in the future?Is there a second-round effect or has this already been

exploited?

Empirical studies necessarily have to rely onhistorical data. New services and technolo-gies such as mobile broadband or Ethernetfor data transfer over shorter distances can-not yet be quantified in terms of their growthcontribution. However, there is good reasonto believe that the effect of ICT will be at leastas great in the future, if not greater. More-over, the literature often points to second-round effects of ICT investment with a cer-tain time delay, which could increase thegrowth contribution still further.

8

Table 2 Overview table: contribution of ICT to labor productivity in %

EU USA ICT

contribution Knowledge economy

contribution

ICT contribution

Knowledge economy

contribution 1990–1995 17 88 57 83 1995–2000 42 78 59–66 81–98 2000–2005 45 67 33–43 70–92 2003–2007 31 89

Sources: USA data: 1990–95: Jorgenson (2001); 1995–2000: Cette (2009), Jorgenson (2001); 2000–05/6: Gordon (2010), Jorgenson (2008), Oliner (2007); EU data: 1990–95: Van Ark (2002); 1995–2000: O'Mahoney and Timmer (2009); 2005–05: EUKLEMS (2008), Van Ark and Inklaar (2005); 2003–07: EC (2011).

Frequency of Elasticities and Elasticities over Time

-0.1

0.0

0.1

0.2

0.3

1980 1985 1990 1995 2000 2005

Source: Authors' calculations.

0

2

4

6

8

-0.1 0.0 0.1 0.2 0.3

Output Elasticity of ICT

Number of Studies

Linear estimation is derived from precision-weighted observations.

ICT Elasticity

Year

Figure 8

Table 3 Elasticities by aggregation level and ICT product (number of studies in brackets)

ICT product/ aggregation level

Company level

Industry level

Country level

Voice telephony (mobile/fixed network)

– – 0.16 (1)

Dates (Internet/broadband)

– – 0.045–0.16 (3)

IT (hardware/software)

0.015–0.39 (15) –0.071–0.17 (2) –

ICT 0.049–0.15 (3) 0.031–0.066 (2) –0.013–0.138 (2)

Source: Studies consulted in the Appendix.

ICT and productivity

Literature

Acharya, Ram; Basu, Susanto (2010): ICT and Total Factor ProductivityGrowth: Intangible Capital or Productive Externalities?Aghion, P.; Howitt, P. (2007): Capital, innovation, and growth accounting. In Oxford Review of Economic Policy 23 (1), pp. 79–93.Baily, Martin N.; Lawrence, Robert Z. (2001): Do We Have a New E-Cono-my? In American Economic Review 91 (2), pp. 308–312.Basu, Susanto; Fernald, John G.; Oulton, Nicholas; Srinivasan, Sylaja (2003):The Case of the Missing Productivity Growth, or Does Information Technol-ogy Explain Why Productivity Accelerated in the United States but Not in theUnited Kingdom? In NBER Macroeconomics Annual 18.Bertschek, Irene; Kaiser, Ulrich (2004): Productivity Effects of Organization-al Change: Microeconometric Evidence. In Management Science 50 (3), pp. 394–404.Bloom, Nicholas; Sadun, Raffaella; van Reenen, John (2010): Americans Do I.T. Better: US Multinationals and the Productivity Miracle.Bresnahan, Timothy F.; Brynjolfsson, Erik; Hitt, Lorin M. (2002): InformationTechnology, Workplace Organization, and the Demand for skilled Labour:Firm-Level Evidence. In The Quarterly Journal of Economics 117 (1), pp. 339–376.Bresnahan, T.F. and M. Trajtenberg (1995), General Purpose Technologies:Engines of Growth? In Journal of Econometrics 65 (1), pp. 83–108.Brynjolfsson, Erik; Hitt, Lorin M. (1995): Information Technology as a Factorof Production: The Role of Differences Among Firms. In Economics of Inno-vation and New Technology 3 (3), pp. 183–200.Brynjolfsson, Erik; Hitt, Lorin M. (1996): Paradox Lost? Firm-Level Evidenceon the Returns to Information Systems Spending. In Management Science 42 (4), pp. 541–558.Brynjolfsson, Erik; Hitt, Lorin M. (2003): Computing Productivity: Firm-LevelEvidence. In The Review of Economics and Statistics 85 (4).Cette, Gilbert; Kocoglu, Yusuf; Mairesse, Jacques (2009): Productivity Growthand Levels in France, Japan, the United Kingdom and the United States inthe Twentieth Century. In National Bureau of Economic Research WorkingPaper Series No. 15577.Correa, Lisa (2006): The economic impact of telecommunications diffusionon UK productivity growth. In Information Economics and Policy 18 (4), pp. 385–404.Conference Board (2011), The Conference Board – Total Economy Database,September 2011.Crandall, Robert W.; Lehr, William; Litan, Robert (2007): The Effects of Broad-band Deployment on Output and Employment: A Cross-sectional Analysisof U.S. Data.Czernich, Nina; Falck, Oliver; Kretschmer, Tobias; Woessmann, Ludger (2011):Broadband Infrastructure and Economic Growth*. In The Economic Jour-nal 121 (552), pp. 505–532.Gordon, Robert J. (2000): Does the »New Economy« Measure up to theGreat Inventions of the Past? In Journal of Economic Perspectives 14 (4),pp. 49–74.Gordon, Robert J. (2010): Revisiting U. S. Productivity Growth over the PastCentury with a View of the Future. In National Bureau of Economic ResearchWorking Paper Series No. 15834.Greenstein, Shane; McDevitt, Ryan C. (2009): The Broadband Bonus: Accounting for Broadband Internet's Impact on U.S. GDP. In National Bureau of Economic Research Working Paper Series No. 14758.Greenstein, Shane; Spiller, Pablo T. (1995): Modern Telecommunications Infrastructure and Economic Activity: An Empirical Investigation. In Industri-al and Corporate Change 4 (4), pp. 647–665.Grimes, Arthur; Ren, Cleo (2009): The Need for Speed: Impacts of InternetConnectivity on Firm Productivity.Hardy, Andrew P. (1980): The role of the telephone in economic development.In Telecommunications Policy 4 (4), pp. 278–286.Hausman, Jerry A.; Pakes, Ariel; Rosston, Gregory L. (1997): Valuing theEffect of Regulation on New Services in Telecommunications. In BrookingsPapers on Economic Activity 1997 (Special Issue Microeconomics), pp.1–54.Hempell, Thomas (2005a): Does Experience Matter? Innovations and theProductivity of Information and Communication Technologies in GermanServices. In Economics of Innovation and New Technology 14 (4), pp. 277–303.Hempell, Thomas (2005b): What’s Spurious, What’s Real? Measuring theProductivity Impacts of ICT at the Firm-Level. In Empirical Economics 30 (2),pp. 427–464.Hempell, Thomas; van Leeuwen, George; van der Wiel, Henry (2004): ICT, Innovation and Business Performance in Services: Evidence for Germanyand the Netherlands.

Im, Kun Shin; Dow, Kevin E.; Grover, Varun (2001): Research Report: A Reexamination of IT Investment and the Market Value of the Firm — An Event Study Methodology. In Information Systems Research 12 (1), pp. 103–117.Inklaar, Robert; O'Mahony, Mary; Robinson, Catherine; Timmer, Marcel P.(2003): Productivity and Competitiveness in the EU and the US. In Mary O'Ma-hony, Bart van Ark (Eds.): EU Productivity and Competitiveness: An IndustryPerspective. Can Europe Resume the Catching-up Process? Luxembourg:Office for Official Publications of the European Communities, pp. 73–148.Inklaar, Robert; Timmer, Marcel P.; van Ark, Bart (2008): Market servicesproductivity across Europe and the US. In Economic Policy 23 (53), pp. 139–194.Jorgenson, Dale W. (2001): Information Technology and the U.S. Economy.In American Economic Review 91 (1), pp. 1–32.Jorgenson, Dale W. (2007): Industry Origins of the American ProductivityResurgence. In Economic Systems Research 19 (3), pp. 229–252.Jorgenson, Dale W.; Ho, Mun S.; Stiroh, Kevin J. (2008): A RetrospectiveLook at the U.S. Productivity Growth Resurgence. In Journal of EconomicPerspectives 22 (1), pp. 3–24.Jorgenson, D.W., M.S. Ho and K.J. Stiroh (2005), Productivity – InformationTechnology and the American Growth Resurgence, 3, Cambridge: MIT Press.Jorgenson, Dale W.; Stiroh, Kevin J. (1999): Information Technology andGrowth. In American Economic Review 89 (2), pp. 109–115.Mefford, Robert N. (1986): Introducing Management into the Production Func-tion. In The Review of Economics and Statistics 68 (1), pp. 96–104.Koutroumpis, Pantelis (2009): The economic impact of broadband ongrowth: A simultaneous approach. In Telecommunications Policy 33 (9),pp. 471–485.Lehr, William; Osorio, Carlos A.; Gillett, Sharon E.; Sirbu, Marvin A. (2006):Measuring Broadband’s Economic Impact.Morrison, Catherine J. (1997): Assessing the Productivity of Information Tech-nology Equipment in U.S. Manufacturing Industries. In The Review of Economics and Statistics 97 (3), pp. 471–481.Oliner, Stephen D.; Sichel, Daniel E. (2000): The Resurgence of Growth inthe Late 1990s: Is Information Technology the Story? In Journal of Econom-ic Perspectives 14 (4), pp. 3–22.Oliner, Stephen D.; Sichel, Daniel E.; Stiroh, Kevin J. (2007): Explaining a Productive Decade. In Brookings Papers on Economic Activity 1, pp. 81–137.Röller, Lars-Hendrik; Waverman, Leonard (2001): Telecommunications Infrastructure and Economic Development: A Simultaneous Approach. InAmerican Economic Review 91 (4), pp. 909–923.Schreyer, P. and D. Pilat (2001): Measuring Productivity, OECD EconomicStudies 33 (2), pp. 127-170.Stiroh, Kevin J. (1998): Computers, Productivity, and Input Substitution. In Economic Inquiry 36 (2), pp. 175–191.Stiroh, Kevin J. (2002a): Are ICT Spillovers Driving the New Economy? In Review of Income and Wealth 48 (1), pp. 33–57.Stiroh, Kevin J. (2002b): Information Technology and the U.S. ProductivityRevival: What Do the Industry Data Say? In American Economic Review 92 (5),pp. 1559–1576.Tam, Kar Yan (1998): The Impact of Information Technology Investments onFirm Performance and Evaluation: Evidence from Newly IndustrializedEconomies. In Information Systems Research 9 (1), pp. 85–98.Tambe, Prasanna; Hitt, Lorin M. (2011): The Productivity of InformationTechnology Investments: New Evidence from IT Labor Data. In SSRN eLi-brary.Timmer, M., G. Ypma and B. van Ark (2003), ICT in the European Union: Dri-ving Productivity Divergence?, Groningen Growth and Development Center,Working Paper GD-67.Van Ark, B., R. Inklaar and R.H. McGuckin (2003), ICT and Productivity in Europe and the United States: Where Do the Differences Come From?, In CESifo Economic Studies 49 (3), pp. 295–318.Van Ark, B., J. Melka, N. Mulder, M.P. Timmer and G. Ypma (2002), ICT Investment and Growth Accounts for the European Union, 1980–2000,Groningen Growth and Development Center, Research Memorandum GD-56.

9

ICT and productivity10

Table A.1 Summary of elasticities

Original author Year publ. Elasticity Unit Dates Region No. obs./yr Start End Black 2001 0.050 Company 1987 1993 AM 638 Black 2004 0.296 Company 1993 1996 AM 284 Bresnahan 2002 0.035 Company 1987 1994 AM 300 Brynjolfsson 1996 0.044 Company 1987 1991 AM 702 Brynjolfsson 1995 0.052 Company 1988 1992 AM n.a. Brynjolfsson 2003 0.058 Company 1987 1994 AM 1,324 Dewan 1997 0.090 Company 1988 1992 AM 773 Gilchrist 2001 0.021 Company 1986 1993 AM 580 Hitt 1996 0.048 Company 1988 1992 AM 370 Lichtenberg 1995 0.098 Company 1988 1991 AM 1,315 Tambe 2011 0.041 Company 1987 2006 AM 1,800 Bertschek 2004 0.152 Company 2000 2000 EU 212 Bloom 2010 0.015 Company 1995 2003 EU 4,809 Hempell 2004 0.041 Company 1996 1998 EU 972 Hempell2 2005 0.060 Company 1994 1999 EU 1,177 Mahr 2010 0.130 Company 2000 2008 EU 182 Hempell1 2005 0.049 Company 1994 1999 EU 1,222 Loveman 1994 – 0.060 Company 1978 1984 WW 60 Basant 2006 0.115 Company 2003 2003 AS 266 McGuckin 2002 0.170 Industry 1980 1996 AM 10 Stiroh 2002 – 0.071 Industry 1973 1999 AM 18 Acharya 2010 0.031 Industry 1973 2004 WW 384 Omahoney 2005 0.066 Industry 1976 2000 WW 55 Venturini 2009 0.138 Country 1980 2004 EU 15 Dewan 2000 – 0.013 Country 1985 1993 WW 36 Koutroumpis 2009 0.012 Country 2002 2007 WW 22 Madden 2000 0.162 Country 1975 1990 WW 43 Röller 2001 0.045 Country 1970 1990 WW 21 Sridhar 2007 0.150 Country 1990 2001 WW 63 Note: Region = AM: America, AS: Asia, WW: Worldwide.

Appendix

ICT and productivity 11

Authors of this study Prof. Dr. Tobias Kretschmer

Tobias Kretschmer studied Business Administration at the University of St. Gallen and took a PhD in Economics from the London Business School. Before his professorship, he was employed as a post-doc at INSEAD (2000–2001), Fontainebleau, and as an associate professor at the London School of Economics (2001–2006).

In 2006 he accepted an invitation to take up the Chair of Strategy, Technology and Organisation at Ludwig-Maximilian University in Mu-nich. Alongside his work at LMU, he has headed the department of Industrial Organisation and New Technologies at the Ifo Institute,

Munich since October 2010.

His research interests and specialisms lie in the strategy and economics of high-tech industries and research into the productivity of individual companies and sectors. Mag.a Mélisande Cardona, MBR

Mélisande Cardona studied Economics at the Vienna University of Economics and Business and the University of Buenos Aires. After her studies, she worked for the Austrian regulatory authority for telecom-munications, where she was particularly involved in analyzing compe-tition in the markets for mobile telephony and broadband Internet. She has been studying for a PhD at the Institute for Strategy, Technology and Organization at Ludwig-Maximilian University in Munich since 2007.

Ms. Cardona is researching the adoption and diffusion of information and communications technology (ICT). A focal area is the empirical study of spill-over effects from the use of ICT. Dr. Thomas Strobel

Thomas Strobel is a member of the department for International Institutional Comparison and Industrial Economics and New Techno-logies at the Ifo Institute. He studied Economics at the University of Konstanz and Ludwig-Maximilian University in Munich. In 2009, he received his PhD in Economics from Ludwig-Maximilian University.

His research interests and specialisms lie in the area of productivity analysis and empirical growth studies. He is especially concerned with the influence of information and communications technology (ICT) and

skill-biased technological change on productivity growth in different industries and with sector-based structural change.